An Introductory Course in Game Theory

October 12, 2012

Contents

Foreward ii

1 Introduction 1
1.1 What is Game Theory? . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 What is a “Game”? . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 The Prisoner’s Dilemma Game . . . . . . . . . . . . . . . . . . . . 4

2 Two-Person Zero-Sum Games 6

2.1 Definition and Representation of Zero-Sum Games . . . . . . . . . 6
2.2 Dominant and Dominated Strategies . . . . . . . . . . . . . . . . 8
2.3 Saddle Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.4 The Minimax Theorem and Mixed Strategies . . . . . . . . . . . . 15
2.5 Graphing Zero-Sum Games . . . . . . . . . . . . . . . . . . . . . . 19

3 Two-Person Nonzero-Sum Games 23

3.1 Nash Equilibria in Pure Strategies . . . . . . . . . . . . . . . . . . 23
3.2 Nash Equilibria in Mixed Strategies . . . . . . . . . . . . . . . . . 24
3.3 Pareto Optimal Outcomes . . . . . . . . . . . . . . . . . . . . . . 27

Glossary 30

Annotated Bibliography 33

i
Foreward

The following lessons provide a good introduction to game theory. They are di-
rected towards students in grades 11 or 12, although may be used as enrichment
as early as grades 9 or 10. The lessons include explanations and activities. The
activities can be used either as classroom activities, as exercises to be completed
in pairs, or as individual exercises. Some activities require that the students work
in pairs or small groups and will be described as such.

The lessons do not contain an exhaustive number of student exercises. Rather,

they are meant to be completed during class time to allow for discussion among
students and as a class. Some of the references listed in the Annotated Bibliog-
raphy contain further exercises should more be required.

A glossary is provided at the end of the lessons for student reference as game
theory introduces a lot of new terminology.

Text written in red appears only in the teacher’s edition.

ii
Chapter 1

Introduction

1.1 What is Game Theory?

Have you ever tried to anticipate someone else’s decision before making your own
so that you profit best? If so, you already know a bit about what game theory is
all about.

Game theory involves choosing the best strategy in a situation where there are
two or more people, or groups of people, making decisions. The result does not
just depend on what action you choose, but also on the actions of your opponents.
To illustrate this, consider a simple game you have probably played before: Rock-
Paper-Scissors. Remember that the rules are “rock breaks scissors”, “scissors cut
paper” and “paper covers rock”. Whether you win or not depends completely on
your opponent’s choice. If you choose “rock” and your opponent chooses “scis-
sors”, then you win; however, if they choose “paper” instead, then you lose. Not
all games have outcomes like these that are strictly win or lose. In many games,
you can often strategize in such a way that you guarantee yourself a certain result
which may provide a sure win over your opponent or, at least, minimize your
losses. In this course, we will look at how this is done.

We hear of war games, mind games, political games, marketing games and adver-
tising games, all of which require decision-making designed to put the principal
player in the best position for winning. In game theory we try to create mathe-
matical models to simulate these games. While the models discussed herein may
sometimes appear to be overly simplified, other game models used in business or
politics, for example, can be very complex. Fortunately they can be translated
into computer programs, allowing simulations of the game-play and strategy tri-
als to be carried out. From there, if proven successful, they are put into practice
where their use becomes a reality.

1
CHAPTER 1. INTRODUCTION 2

1.2 What is a “Game”?

Games can be broken down into categories, some of which include games of chance,
games of physical skill, games of mental skill, and games requiring combinations
of chance and skill. The main goal of any game is to win. Winning is usually
measured by keeping score, maximizing it as in basketball or poker, or minimizing
it as in golf. This course will concentrate on games involving mental skills where
both you and the opponent are making decisions.

It is important that you understand that the ideas behind game theory are not
just restricted to simple games such as Rock-Paper-Scissors. Many of the games
we will examine are metaphors for real-life situations in business, politics and
other social interactions. These all involve “games” in a different sense that you
may be used to, but there are common elements which help us to define what a
game really is.

A game must have two or more players. These players can be individuals, a com-
pany, a political party, etc. These players must have a certain set of actions from
which to choose, called strategies. The result from each player picking a certain
strategy gives an outcome and this outcome has certain payoffs to each player.
Payoffs are often represented by either positive or negative numbers, where posi-
tive means a gain and negative means a loss.

The players, strategies, outcomes and payoffs of a game are represented in a ma-
trix, or table. Examine the following game involving two players, Red (the row
player) and Blue (the column player):

Blue
X Y
A (5, -5) (-3, 2)
Red B (2, -1) (0, 0)
C (-3, 3) (5, -4)

Game 1.2.1

Red has three strategies (A, B and C) while Blue has two strategies (X and Y).
For each outcome in a game, the first number always represents the payoff to the
row player (Red) while the second number represents the payoff to the column
player (Blue). Thus, if Red chooses strategy C and Blue chooses strategy Y, Red
will gain 5 while Blue loses 4.

For our purposes, we are going to keep things simple. Assume that the games we
will analyze involve rational players. This means that each player wants to keep
his or her own best interests in mind and ignores how their decision may affect
the other players. We will also assume that there is no communication between
CHAPTER 1. INTRODUCTION 3

players; that is, each player does not know what strategy the other is choosing.
We will further assume that the players either make their choices simultaneously
or, if not, they are not aware of what choice the other player has made before
making their own. In real life, these restrictions are not always there. In fact,
game theory can get quite complicated when you factor in potential cooperation
between opponents or “insider” information which may shed light on an oppo-
nent’s strategy.

Activity

This activity allows for student practice in reading a game matrix. A firm grasp of
this skill is necessary before other lessons can be understood. It is recommended
that the questions be completed in pairs and that a class discussion follows.

1. Examine the following game matrix:

Greg
Red Yellow Green Blue
Alice Heads (3, -8) (-3, -1) (2, 4) (-10, 5)
Tails (2, -1) (0, 0) (4, 6) (2, -7)

a) Describe the game that Alice and Greg could be playing. Remember that
in the games we are studying, the players are choosing their strategies, not
leaving it up to chance, and making their choices without knowledge of what
their opponent’s choice will be.
Answers may vary. One option is that Alice could be choosing to show either
‘Heads’ or ‘Tails’ on a coin, while Greg could be choosing to show one of
four chips coloured red, yellow, green and blue. Depending on the choices
of each player, the payoffs to each vary.

b) How many strategies are available to each player in this game?

Alice has two strategies to choose from while Greg has four.

c) What is the outcome if Alice chooses ‘Heads’ and Greg chooses ‘Blue’ ?
What is the payoff to each player?
The outcome is (-10, 5). Alice’s payoff is -10 (lose 10 points, for example)
while Greg’s payoff is 5 (gain 5 points).

d) If Alice chooses ‘Tails’, which of Greg’s strategies would give him the worst
payoff? The best payoff?
‘Blue’ would give Greg the worse payoff, since -7 is the lowest second number
in each of the entries in the ‘Tails’ row. ‘Green’ would give him the best
payoff, since 6 is the highest second number in each of the entries in the
‘Tails’ row.
CHAPTER 1. INTRODUCTION 4

1.3 The Prisoner’s Dilemma Game

Let us start by looking at a game central to game theory, called the Prisoner’s
Dilemma. Two suspects are arrested and taken in for separate questioning for a
joint crime, but the police do not have enough to convict either of them. Here,
the suspects have only two strategies: stay silent or rat out (implicate) the other.
If a suspect implicates the other while the other stays silent, the suspect who
implicates gets a sentence of 6 months while the other gets a sentence of 10 years
in prison. If both remain silent, they each receive a 1 year sentence, and if both
implicate each other, they are sentenced to 5 years in prison. As we mentioned
before, we are to assume that the suspects cannot communicate with each other
so have no knowledge of the other’s strategy. They are also going to choose ra-
tionally, which again means that they will keep their own best interests in mind.

Activity

It is suggested that the following activity be completed in pairs. This allows for
discussion among students. Once completed, a class discussion should follow.

1. Create a matrix for the Prisoner’s Dilemma game. This type of game is said
to be a symmetric game, since switching the labels for the row player and
column player does not affect each individual’s payoffs.
Individual matrices may vary in their placement of the strategies, but should
have all information as in the following:

Suspect 2
Stay Silent Implicate
Suspect 1 Stay Silent (1 yr, 1 yr) (10 yr, 6 mo)
Implicate (6 mo, 10 yr) (5 yr, 5 yr)

2. Examine the payoffs for Suspect 1. In each case of Suspect 2 choosing to stay
silent or implicate, would Suspect 1 do better to stay silent or to implicate?
Explain.
If Suspect 2 stays silent, Suspect 1 is better off to implicate since a 6 month
sentence is less than a 1 year sentence. If Suspect 2 implicates, Suspect 1 is
again better off implicating since 5 years is a lower sentence than 10 years.
In either case, Suspect 1 is better off to implicate.

3. Now examine the payoffs for Suspect 2. In each case of Suspect 1 choosing
to stay silent or implicate, would Suspect 2 do better to stay silent or to
implicate? Explain.
The same reasoning applies here since Suspect 2 has the same payoffs as
Suspect 1, so Suspect 2 should choose to implicate in either case.
CHAPTER 1. INTRODUCTION 5

4. Assuming that both suspects make their decision based on the reasoning in
questions 1 and 2, what will the outcome be? Is this the best outcome for
each suspect? Explain.
Both suspects would receive a 5 year sentence. This is definitely not the
best outcome. Obviously, it would be better to receive a 6 month sentence
or even a 1 year sentence, but it is impossible to determine what the other
suspect would choose. The best scenario for both would be if both stayed
silent; however, there is no way to guarantee that the other suspect would
choose this strategy. Therefore, given the circumstances, the safest choice
would be to implicate and unfortunately, receive a 5 year sentence.

The Prisoner’s Dilemma is an example of a game where the likely outcome is not
optimal for either player, but it is the safest. Not all games are of this nature, as
we will see. It is also not a game that involves repetition, or ”rounds”, such as in
Rock-Paper-Scissors, where you might repeat three times to decide a winner. In
the following chapters, you will learn how to determine the safest outcome for a
number of different games, and in doing so ensure that you have made the best
decision in the given circumstances.

Activity

The following may be completed individually or in small groups.

1. Invent a scenario that could correspond to the following game matrix. Be

sure to specify who the players might be and explain why the payoffs might
be as they are.
Player 2
Ballet Ball Game
Player 1 Ballet (1, 5) (0, 0)
Ball Game (0, 0) (5, 1)

Answers may vary. One possibility is a husband (Player 1) and wife (Player
2) are trying to agree on how to spend the evening. The husband prefers
going to a ball game while the wife would rather attend the ballet. If they
disagree, no one is happy, so the payoff to both is zero. If the wife gets
her preference but the husband doesn’t, the wife is elated so her payoff is 5
while the husband is unhappy (but not as much as if they disagreed) so his
payoff is 1. A similar situation occurs when the husband gets his preference
but the wife doesn’t.
Chapter 2

Two-Person Zero-Sum Games

2.1 Definition and Representation of Zero-Sum

Games
Some games in game theory are such that the players are in pure conflict: what-
ever the Red player gains, the Blue player loses and visa versa. These types of
games are called zero-sum, since the payoffs for each outcome add to zero. Take,
for example, the following zero-sum game:

Blue
A B C
A (5, -5) (-3, 3) (-2, 2)
Red B (4, -4) (1, -1) (2, -2)
C (3, -3) (-2, 2) (4, -4)

Game 2.1.1

For simplicity, zero-sum games are often represented in such a way that only
the payoffs for the Red player are shown as it is implied that the Blue player’s
payoff is the negation. Game 2.1.1 can be represented by the much simpler matrix:

Blue
A B C
A 5 -3 -2
Red B 4 1 2
C 3 -2 4

Game 2.1.1

6
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 7

Activity

The following activity allows students to practice reading the payoffs in a zero-
sum game, where only the row player’s payoffs are shown and the column player’s
payoffs are the corresponding negative. Ensure that students are comfortable with
the activity before moving on.

1. Examine the following game matrix:

Blue
X Y Z
A 2 0 -6
Red B -1 10 -4
C 5 7 -3

a) What is the payoff to Red if Red chooses B while Blue chooses Y?

10

b) What is the payoff to Blue if Red chooses C while Blue chooses X? What if
Red chooses C while Blue chooses Z?
-5 and 3

c) Which choices for each player would result in the worst possible payoff to
Red? What about the best possible payoff? Explain.
Red would be worst off if Red A - Blue Z were the outcome since this is the
lowest entry in the matrix, which corresponds to the lowest payoff to Red.
Red would be best off if Red B - Blue Y were the outcome since this is the
highest entry in the matrix, which corresponds to the highest payoff to Red.

d) Which choices for each player would result in the worst possible payoff to
Blue? What about the best possible payoff? Explain.
Blue would be worst off if Red B - Blue Y were the outcome since this is the
highest entry in the matrix, which corresponds to the lowest payoff to Blue.
Blue would be best off if Red A - Blue Z were the outcome since this is the
lowest entry in the matrix, which corresponds to the highest payoff to Blue.

Activity
1. Play 20 rounds of Rock-Paper-Scissors with a partner. Assume that when
a player wins a round, he gains a point while the other player loses a point.
Use the following chart to keep a tally of what each player chose.
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 8

Player 2
Rock Paper Scissors

Rock

Player 1 Paper

Scissors

a) Which combinations of strategies result in neither player gaining a point?

Rock - Rock, Paper - Paper and Scissors - Scissors.

b) Which combinations of strageties result in Player 1 gaining a point while

Player 2 loses a point? (Always write the row player’s strategy first, followed
by the column player’s strategy.)
Paper - Rock, Scissors - Paper and Rock - Scissors.

c) Which combinations of strategies result in Player 1 losing a point while

Player 2 gains a point?
Rock - Paper, Paper - Scissors and Scissors - Rock.

d) Create the game matrix for Rock-Paper-Scissors, played by a Red and a

Blue player. Represent the game in terms of payoffs to the Red player.
Placement of the strategies may vary. A suitable matrix would be:

Blue
Rock Paper Scissors
Rock 0 -1 1
Red Paper 1 0 -1
Scissors -1 1 0

2.2 Dominant and Dominated Strategies

Activity

With a partner, play 20 rounds of the following game. Decide who will play as
Red and who will play as Blue. Each person should record their strategy choice
and the payoff they received in each round. Remember that the Red player wants
large payoffs while the Blue player wants small payoffs, since his payoffs are the
corresponding negative of Red’s payoffs.
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 9

Blue
A B C
A 5 -3 -2
Red B 4 1 2
C 3 -2 4

After giving students the opportunity to play the game, have them try to answer
the following questions. A class discussion should follow.
1. Of Blue’s three strategies, which is the best? Explain why.
Strategy B is the best for the Blue player because it contains the lowest
numbers i.e. the highest payoffs to Blue.
2. Is there a distinct “best” strategy for the Red player? Explain.
Strategy B for the Red player contains only positive numbers, which cor-
respond to gains; however, strategy A gives the potential to achieve the
highest payoff. Therefore, there is no distinct “best” strategy for Red.
3. Is there a strategy that the Blue player should avoid completely?
Blue should avoid playing A, since there are no negative numbers in this
column and so he can only lose points. Also, this column contains Red’s
highest payoff, which he will want to avoid.
4. If the Red player knows that Blue will avoid the strategy from question 3,
is there a strategy that the Red player should avoid?
If Red determines that Blue will never play his strategy A, Red should avoid
his strategy A since the only remaining payoffs in that row are -3 and -2,
both losses.
In the game you just played, you should have found that the Blue player’s best
strategy is to play strategy B. In fact, there is no way that Blue can do any better
by playing any other strategy. We say that Blue B dominates both Blue A and
Blue C, or that Blue A and Blue C are dominated by Blue B. Blue B is called the
dominant strategy while Blue A and Blue C are dominated strategies.

Since the players are playing rationally, and they each know this, dominated
strategies can be removed from the matrix. This is known as the Dominance
Principle: no player will play a dominated strategy. It is possible that a strategy
that was not dominated at first will become dominated and can subsequently be
removed from the game. The result is a new smaller and simpler matrix which is
easier to analyze.

The process of eliminating dominated strategies requires careful observation of

the game matrix. We will use the following game to illustrate this process:
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 10

Blue
A B C D
A 3 -6 2 -4
Red B 2 1 0 1
C 2 3 1 4

Game 2.2.1

First, look at the payoffs to Blue under each of his available strategies. Notice that
column B’s values are all lower than those in column D. Thus, we may remove
strategy D since it is dominated. Similarly, the values in column C are lower than
those in column A, so strategy A may be removed. Blue would not want to play
either of these strategies since they would always result in lower payoffs. Removal
of these strategies gives the following game matrix:

Blue
B C
A -6 2
Red B 1 0
C 3 1

Game 2.2.1 - Blue’s dominated strategies removed

There are no remaining columns with strictly lower values than another, so we
have removed all of Blue’s dominated strategies. Now examine the payoffs to Red
beside each of his available strategies. Row C’s values are all higher than those
in row B, so we may remove strategy B since it is dominated. The game matrix
now becomes:
Blue
B C
Red A -6 2
C 3 1

Game 2.2.1 - Red’s dominated strategy removed

There are no remaining rows with strictly higher values than another, so we have
removed all of Red’s dominated strategies. The result is a much simpler game
which will be easier to solve, as we will see in the following sections.
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 11

Activity
1. For the following game, remove any dominated strategies. Continue to do
so until there is are no more such strategies. Which strategies remain for
each player? Draw the resulting matrix.

Blue
A B C D
A 3 -2 1 2
Red B 4 1 2 3
C 0 0 -3 -2
D -2 -4 5 4

Game 2.2.2

The order in which strategies are removed may vary, but the final result
should be the same. A sample answer is as follows: Blue A and C are both
dominated by B, so they should both be removed. Red A and C are both
dominated by Red B so they should be removed. After this result, Blue D
is dominated by B so should be removed. Finally, Red D is dominated by
Red B, so Red D should be removed. The only remaining strategies are Red
B and Blue B:
Blue
B
Red B 1

2. What is the outcome of the game? Is the game a fair game? Explain.
The outcome is that Red will gain 1 and Blue will lose 1. This is not a fair
game, since Red will always win when the game is played rationally.

2.3 Saddle Points

Game 2.2.2 was able to be reduced down to a single outcome. We call this out-
come the equilibrium outcome or a saddle point, and it is also the solution
to the game. A saddle point is an outcome in a zero-sum game that is simulta-
neously smallest in its row and largest in its column. There are two ways to find
a saddle point, and both ways are faster than eliminating dominated strategies.
Some games have more than one saddle point, and we will see how to deal with
these types in the next section.

The first way to determine if there is a saddle point or not is what we will call the
maximin-minimax method. First, find the minimum entry in each of Red’s
rows and from these choose the highest entry, called the maximin (the maxi mum
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 12

minimum). Then find the maximum entry in each of Blue’s columns and choose
the lowest entry, called the minimax (the mini mum max imum). If the minimax
and maximin have the same value, this is a saddle point and is the solution to the
game. If the minimax and maximin do not have the same value, you must resort
to another method to find a solution. This will be discussed in the next section.

Here are the row minima and column maxima for Game 2.2.2:

Blue
A B C D Row min.
A 3 -2 1 2 -2
Red B 4 1 2 3 1 ← maximin
C 0 0 -3 -2 -3
D -2 -4 5 4 -4

Column max. 4 1 5 4
↑
minimax
The maximin and minimax both have value 1, so Red B - Blue B is a saddle
point and is therefore the solution to the game. Both players are best off choosing
these strategies. We say the value of the game is 1. On average, you cannot
win more than the value of the game unless your opponent is playing irrationally.
Unfortunately for the Blue player, the game is unfair since on average he will lose
1 while Red gains 1.

Why does this method provide a solution? Red wants to maximize his minimum
payoff so that he is guaranteed not to get less than this value. At the same time,
Blue wants to minimize Red’s maximum payoff. Therefore, Blue wants to choose
the minimum of his column entries to ensure that Red’s payoff will be at most
this value. This also ensures that he will lose no more than this value.

A saddle point represents an equilibrium strategy and each player does best by
choosing the strategy that corresponds to it. If a player deviates from this par-
ticular strategy, he will do worse. If both players deviate from their saddle point
strategies, one or both will do worse.

Sometimes, you may have two same numbers that are row minima or column
maxima. This is okay and just means that there are multiple saddle points. It
is interesting to note that if a game has multiple saddle points, they will always
contain the same payoffs.

Another way to find saddle points is to use what we will call the circle-box
method. It is the same idea as the maximin-minimax method, but is quicker and
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 13

more visual. Here, the game is examined from each player’s point of view. For
each of Blue’s strategies, the Red player would want to have the greatest payoff, so
for each column we circle the largest entry. For each of Red’s strategies, the Blue
player would also want the highest payoff (the lowest number), so for each row
we box the smallest entry. If there is an entry in the matrix which has been both
circled and boxed, then this entry is a saddle point and is the solution to the game.

In Game 2.3.1, the largest entry in each column has been circled and the smallest
entry in each row has been boxed:

Game 2.3.1

Since the entry at Red C - Blue B is both circled and boxed, this is a saddle point
and the value of the game is 2.

Activity

Adapted from Williams (1986, p.24)

1. A husband and wife are planning a camping trip. The husband likes high
altitudes, while the wife prefers low altitudes. They only thing they have
agreed on so far is to camp at a path junction, where the husband chooses
the east-west path and the wife the north-south path. Each has a choice
of 4 paths. They have decided to write their choices in secret on a slip of
paper and will simultaneously show one another. The outcome will decide
their camping spot. The following matrix shows the altitudes, in thousands
of feet, at each of the possible junctions.

Assume that the paths cross each other as shown in the matrix. For sim-
plicity, we are assuming the paths are straight, although they are truly not!
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 14

a) Without regard for the wife’s possible choices, which junction would the
husband prefer most? Which would the wife prefer most?
The husband would prefer AW, since it corresponds to 7000 feet, the highest
available altitude. The wife would prefer either DY or AZ, since either
corresponds to the lowest possible altitude of 1000 feet.

b) Find the solution to the game using either the maximin-minimax method
or the circle-box method. Which paths should therefore be chosen by the
husband and wife?
Wife’s Choices
W X Y Z Row min.
A 7 2 5 1 1
Husband’s B 2 2 3 4 2
Choices C 5 3 4 4 3 ← maximin
D 3 2 1 6 1

Column max. 7 3 5 6
↑
minimax
or

The solution is CX, corresponding to an altitude of 3000 feet.

c) Explain why if the husband stays with his corresponding saddle point strat-
egy but the wife deviates from hers, she will only do worse.
The husband’s corresponding saddle point strategy is C. If the wife were to
choose a strategy other than X, she will have to camp at an altitude of 5000
feet (CW) or 4000 feet (CY or CZ). This is worse for her than the saddle
point altitude of 3000 feet.

d) Show that if the husband and wife both deviate from their corresponding
saddle point strategies, either the husband or the wife will do the same or
better and the other will be worse off.
Choices of outcomes may vary. For example, if BZ were the result, corre-
sponding to an altitude of 4000 feet, the husband would be happier but the
wife would be unhappier since this is a higher altitude. If AX was the result,
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 15

the wife would be happier but the husband not since this corresponds to an
altitude of only 2000 feet.

2. With a partner, create a game matrix of your own that has a saddle point.
Clearly show where the saddle point is and explain how you found it.
Answers may vary. The teacher or another student group should verify that
the game matrix does in fact have a saddle point.

2.4 The Minimax Theorem and Mixed Strate-

gies
As we mentioned before, some zero-sum games do not have a saddle point. This
does not mean the game does not have a solution. In fact, the brilliant scientist
and mathematician John von Neumann proved the Minimax Theorem which
states that:

• Every two-person zero-sum game has a solution.

• The solution is called the value of the game and is a unique number v such
that the row player’s best possible payoff is v no matter what the column
player does and the column player’s best possible payoff is -v no matter
what the row player does (Straffin, 1993, p.18).

The strategies used in section 2.3 are called pure strategies because the players
do best to select one particular strategy and not deviate from it. For games that
do not have a saddle point, the solution calls for using mixed strategies. This
involves playing several strategies according to certain probabilities. The solution
to a game with no saddle point may involve one player using a pure strategy while
the other uses a mixed strategy or both players using mixed strategies.

At this point, we will only look at 2×2 games with no saddle point, those for
which each of the two players only has two strategies to select from, as they are
much less complicated. In the next section, we will examine how to solve 2×n
or m×2 games with no saddle point, where one player has two strategies while
the other has more. Games that are any larger get quite complicated, though if
you can handle the complicated algebra you can solve them using the techniques
which will be shown.

To find an optimal mixed strategy, we need to be able to find an expected payoff.

An expected payoff is calculated by multiplying each payoff by its associated
probability and adding them all up. For example, if you were playing a game
with a six-sided die for which you receive 10 points if you roll a 6 and lose 4
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 16

points otherwise, then the expected payoff is

1 5 10 20 10 5
   
10 + (−4) = − = − = − = −1.6
6 6 6 6 6 3
In other words, you would lose, on average, 1.6 points.

To find the optimal mixed strategy for Red, you must first write an expression
for his expected payoff under each of Blue’s two pure strategies. These expected
payoffs are set equal to one another and the resulting equation is solved to find
the best mixed strategy. In doing this, Red finds the best probabilities with which
to play each strategy that Blue cannot take advantage of. Essentially, Red “im-
munizes” himself against Blue’s strategies. Thus, Red can guarantee himself a
certain payoff no matter what strategy Blue chooses. Similarly, Blue will also
have certain probabilities that should be used with each of his strategies to ensure
he gets a certain payoff. Just like for a saddle point, these mixed strategies are
the best each player can do and if they are deviated from, the players will only
do worse.

To illustrate this process, we will look at a zero-sum battle between two major
television networks, NBC and CBS (adapted from Rosenthal, 2011, p.42). Each
network has decided to run either a sitcom or a sports program during a particular
prime time slot. From past research, the networks both know by how many per-
centage points they will be behind or ahead each other in each of the four possible
cases. The percentage points for NBC are shown in the following game matrix. As
this is a zero-sum game, CBS’s percentage points are the corresponding negative.

CBS
sitcom sports
NBC sitcom 10 -6
sports -8 -4

Game 2.4.1

It is important that you first look for a saddle point or dominant strategies before
you resort to finding mixed strategies. Verify that there are none in Game 2.4.1.

Now we find the expected payoff for NBC under each of CBS’s strategies. Let x
be the probability of NBC running a sitcom and 1 − x be that of NBC running a
sports program. If CBS chooses to play a sitcom, the expected payoff for NBC is

10x + (−8)(1 − x)

and if CBS plays a sports program, the expected payoff for NBC in this case is

−6x + (−4)(1 − x)
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 17

Setting these two expected payoffs to be the same and solving for x gives

10x + (−8)(1 − x) = −6x + (−4)(1 − x)

18x − 8 = −2x − 4
20x = 4
1
x =
5
Therefore, NBC should randomly run a sitcom with probability 15 and a sports
program with probability 1 − 51 = 45 . Substituting x = 1
 5
into the first expression
1
for NBC’s expected payoff, 10x + (−8)(1 − x) = 10 5 + (−8) 45 = − 22 5
= −4.4.
Note that we could have also substituted into −6x + (−4)(1 − x) and obtained
the same result. Unfortunately, this means that, on average, NBC will be 4.4
percentage points behind CBS. This is the best NBC can do.

Looking at CBS’s optimal mixed strategy, let y be the probability of CBS running
a sitcom and 1 − y be that of CBS running a sports program. Again, we set CBS’s
expected payoffs under NBC’s two choices equal to each other and solve. Do not
forget to negate the values in the matrix to get CBS’s payoffs.

−10y + 6(1 − y) = 8y + 4(1 − y)

−16y + 6 = 4y + 4
−20y = −2
1
y =
10
1
So CBS should randomly run a sitcom with probability and a sports program
  10  
9 1 9
with probability 10 . CBS’s expected payoff is −10 10 + 6 10 = 22
5
= 4.4.

It is no coincidence that CBS gains 4.4 while NBC loses 4.4. The Minimax The-
orem told us this will always be the case.

In giving the solution of a mixed strategy game, we state both Red and Blue’s
optimal mixed strategies as well as the value of the game in terms of Red  (the 
expected payoff to Red). Thus, for our network battle, the solution is NBC 15 , 45 ,
 
1 9
CBS ,
10 10
with value − 22
5
or -4.4.
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 18

Activity

Students will work in pairs and each require a coin.

1. A friend approaches you to play the Matching Pennies game. In this game,
each player has a penny that is turned to show either heads or tails. The
players’ choices are revealed simultaneously. Your friend proposes that you
receive 3 points if he chooses tails when you choose heads, 1 point if he
chooses heads while you choose tails and you lose 2 points when the coins
match.
a) Play 20 rounds of the game, keeping track of which strategies each player
chose and their payoffs.
b) Create the game matrix for the Matching Pennies game, showing your pay-
offs.
Placement of strategies may vary. One possible game matrix is
Friend
heads tails
You heads -2 3
tails 1 -2
c) Verify that there is no saddle point and no dominant strategies.
The circle-box method reveals no saddle point. Inspection of the matrix
reveals no dominant strategy for either player.
d) What is the solution to the game? Is the game fair? Explain.
The solution calls for mixed strategies, as there is no saddle point. Equating
your expected payoffs gives
−2x + 1(1 − x) = 3x + (−2)(1 − x)
−3x + 1 = 5x − 2
−8x = −3
3
x =
8
3
so you should play heads with probability 8
and tails with probability 58 .

Equating your friend’s expected payoffs gives

2y + (−3)(1 − y) = −1y + 2(1 − y)
5y − 3 = −3y + 2
8y = 5
5
y =
8
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 19

5
so your friend should play heads with probability 8
and tails with proba-
bility 83 .
   
The value of the game is −2 38 + 1 58 = − 18 = −0.125. On average you
will lose 12.5 points, so the game is not fair. A game is only fair when the
value of the game is zero.

e) Suggest a method you could use to ensure that you randomize

  your mixed
1 1
strategy. For example, if you had a mixed strategy of 2 , 2 , you could flip
a coin to decide which strategy to play.
Answers may vary. One solution is to put 3 slips of paper marked “heads”
and 5 slips of paper marked “tails” in a hat and draw, replacing the papers
each time.

f) Suppose the game was such that the players do not choose whether to display
heads or tails, but leave it up to chance by flipping the coins. Show that
this would make a fair game.
The probability of each of the four possible outcomes (HH, HT, TH, TT) is
1
4
, so the value of the game would be

1 1 1 1
       
−2 +1 +3 + (−2) =0
4 4 4 4

This would now be a fair game.

2.5 Graphing Zero-Sum Games

Zero-sum games that are 2×2 can be solved algebraically as we have seen, but 2×n
or m×2 games where there is no saddle point or dominated strategies to eliminate
require a different approach. By the Minimax Theorem, there is a solution and
it happens that it is the mixed strategy solution to one of the 2×2 subgames. To
find out which 2×2 subgame to analyze, it is helpful to graph the game.

We will solve Game 2.5.1 to show the process, in which Red has 2 strategies to
choose from and Blue has 4.
Blue
A B C D
Red A 2 0 3 -2
B -1 0 -3 1

Game 2.5.1
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 20

We first check for a saddle point and dominant strategies. There are none (ver-
ify this for yourself). To graph the game, we make two vertical axes, one for
each of the two strategies available to the player with two strategies, in this case
Red. For each of Blue’s strategies, a line is drawn from his matrix payoff on the
left axis when Red plays A to his matrix payoff on the right axis when Red plays B.

Graph of Game 2.5.1

We then decide whether the player with multiple strategies wants outcomes that
lie on the upper envelope of the graph, the set of line segments along the top
portion of the graph (highlighted in green), or the lower envelope, the set of line
segments along the lower portion of the graph (highlighted in orange). In our case,
Blue wants to make Red’s payoff as small as possible. Therefore, he would choose
a mixed strategy on the lower envelope and knowing this, Red would choose his
mixed strategy with probability x so that he is on the highest point of the lower
envelope. This point has been circled on the graph, and corresponds to where
Blue C and Blue D intersect. Therefore, the 2×2 subgame we need to solve is

Blue
C D
Red A 3 -2
B -3 1

Appropriate subgame for Game 2.5.1

Using the method of equating expected payoffs described in section 2.4, the so-
lution is that Red should play A with probability 49 and B with probability 59 ,
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 21

written as Red ( 94 , 95 ), while Blue should not play A or B but play C with proba-
bility 13 and D with probability 23 , written as Blue (0, 0, 13 , 23 ). The value of the
game is − 13 , so on average Red will lose 13 while Blue will gain 31 .

Notice that the intersection of Blue C and Blue D is closer to Red B. In fact, it is
5
9
of the way to Red B. This is no coincidence. We can approximate Red’s mixed
strategy from the graph itself. Not only that, we can also estimate the value of the
game by reading the value of the intersection point on the vertical axes. Check
for yourself that the intersection is at − 13 .

Activity

Adapted from Williams (1986, p.76)

Some games are played against nature, for which we have a player that will not
play rationally, but more unpredictably. Game theory can still be applied in this
case, although the following activity simplifies the situation.

1. An amplifier manufacturer uses a specific part that normally costs only $1

but sets them back $10 if the part is defective. A better quality part is
available for $6 which is fully guaranteed; that is, the maker of the part will
replace the part should it be defective. A third type for $10 is also available,
which comes with an insurance policy claiming to cover the cost of the part
and replace it for free should it be defective. Which part should be used in
the amplifier so as to save the most money?

a) Create the matrix for the game in terms of the amplifier manufacturer’s
payoffs. Use negative values to represent money spent.
Part
Defective Not Defective
Cheap part -10 -1
Amplifier Guaranteed part -6 -6
manufacturer Insured part 0 -10

b) Verify that there is no saddle point.

The circle-box method reveals that there is no saddle point.
CHAPTER 2. TWO-PERSON ZERO-SUM GAMES 22

c) Graph the game and determine which 2×2 subgame should be analyzed.

The manufacturer would want an outcome that lies on the upper envelope,
since this corresponds to a larger number and thus a lower amount of money
spent. The strategies that correspond to the lowest point on the upper
envelope involve the ”cheap part” and the ”insured part”.
d) Write the matrix for the reduced game and solve it. That is, find the best
mixed strategy for the amplifier manufacturer and the value of the game.
Part
Defective Not Defective
Amplifier Cheap part -10 -1
manufacturer Insured part 0 -10

Let x be the probability with which a cheap part should be equipped. Equat-
ing expected payoffs gives:
−10x + 0(1 − x) = −1x + −10(1 − x)
−19x = −10
10
x =
19
so the amplifier manufacturer should equip the cheap part with probability
10 9
19
and the more expensive insured part with probability 19 . That is, they
should use 10 cheap parts for every 9 insured parts. The guaranteed parts
should not be used at all.
The value of the game is −10( 10
19
9
) + 0( 19 ) = −$5.26.
e) What does the value of the game mean in this case? Explain.
The amplifier manufacturer will end up spending, on average, $5.26 per part
for the amplifier.
Chapter 3

Two-Person Nonzero-Sum Games

3.1 Nash Equilibria in Pure Strategies

Now we will turn back to games such as the Prisoner’s Dilemma introduced in
section 1.3. This game is called a nonzero-sum game, since the payoffs for each
outcome do not add to zero. As such it is required that in the game matrix we
show the payoff for both Red and Blue. It is convention that we write the row
player’s payoff first, followed by the column player’s payoff.

Some of what we learned about zero-sum games can be applied to nonzero-sum

games. We can look for a pure strategy equilibrium, the equivalent of a saddle
point in a zero-sum game, and if none exists or if there are multiple pure strategy
equilibria that are not equivalent, we can find a mixed strategy. However, some-
times this pure or mixed strategy equilibrium is not the most optimal solution,
so some nonzero-sum games cannot be solved unless we allow for player coopera-
tion or knowledge of the opponent’s choice before making a decision. We will not
examine games of this type as they can become very complicated.

In nonzero-sum games we call any equilibrium outcome a Nash equilibrium,

named after John Nash, a mathematician who made great contributions to game
theory. He proved that every nonzero-sum game has at least one equilibrium and
this equilibrium is obtained from either pure strategies or mixed strategies.

We will first look at a game that has a pure strategy Nash equilibrium, which
can be found by the circle-box method. Remember that both Red and Blue want
to have the greatest payoff for any given opponent strategy, so we will circle the
best payoff in each column among the first entries (those for Red) and box the
best payoff in each row among the second entries (those for Blue). Any outcomes
for which the first payoff is circled and the second payoff is boxed are pure strategy
Nash equilibria.

23
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 24

Game 3.1.1 illustrates how to find a pure strategy Nash equilibrium. Red A -
Blue X is the solution.

Game 3.1.1
Activity
1. Examine the following nonzero-sum game:
Blue
A B C
A (0, -1) (0, 2) (2, -3)
Red B (0, 0) (2, 1) (1, -1)
C (2, 2) (1, 4) (1, -1)
a) Use the circle-box method to determine the Nash equilibrium. Which strate-
gies should Red and Blue therefore choose?
The Nash equilibrium is (2, 1), when both have chosen strategy B.
b) Is this outcome the most desirable to each player? Explain.
The payoff of 2 is the most desirable for Red, since any other outcome gives a
payoff to Red of 2 or less. The payoff of 1 for Blue is not the most desirable,
since there are other strategies that give a higher payoff to Blue, such as
Red A - Blue B, Red C - Blue A or Red C - Blue B.

3.2 Nash Equilibria in Mixed Strategies

Finding the appropriate strategy mix for a nonzero-sum game is similar to the
method of equating expected payoffs we used in section 2.4 but with one important
difference. Instead of Red considering his own payoffs in deciding his mixed strat-
egy, he will need to consider Blue’s payoffs to neutralize Blue’s possible choices.
Similarly, Blue will need to consider Red’s payoffs to neutralize Red’s choices.

We will use the following game to illustrate this concept:

Blue
C D
Red A (4, 8) (2, 0)
B (6, 2) (0, 8)
Game 3.2.1
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 25

There is no pure strategy Nash equilibrium (no equilibrium found from the circle-
box method) so we must find a mixed strategy Nash equilibrium. Red only
needs to consider Blue’s payoffs to determine his mixed strategy, so the matrix he
will examine is:
Blue
C D
Red A 8 0
B 2 8

Let x be the probability of Red choosing strategy A and 1 − x be that of choosing

strategy B. If Blue chooses strategy C, then Red can expect the payoff to Blue to
be
8x + 2(1 − x) (1)
and if Blue chooses strategy D, Blue’s expected payoff is

0x + 8(1 − x) (2)

Equating these and solving gives x = 37 , so Red should play A with probability 3
7
and B with probability 47 .

The matrix that Blue needs to consider is:

Blue
C D
Red A 4 2
B 6 0

Let y be the probability of Blue choosing strategy C and 1 − y be that of choosing

strategy D. Equating expected payoffs to Red as we did above gives

4y + 2(1 − y) = 6y + 0(1 − y) (3)

1
which can be solved to yield y = 2
, so Blue should play C and D with equal
probabilities.

By substituting y = 21 into the left side (or right side,


if you prefer) of equation
(3), the average payoff to Red is then 4 2 + 2 2 = 3. By substituting x = 37
1 1
   
into equation (1) we get the average payoff to Blue as 8 37 + 2 47 = 4 47 . Blue
therefore has a slight advantage in the game, since his payoff is greater.

Some nonzero-sum games have more than one equilibrium in pure strategies. In
this case, find the optimal mixed strategy and the average payoff to each player
as we did above. Sometimes, the resulting average payoffs to each player are less
than those if one of the pure strategy equilibria were chosen, so the game is not
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 26

strictly solvable. This will be discussed further in the next section.

Activity
1. Besides the Prisoner’s Dilemma, another game central to game theory is
called the Stag Hunt. In the game, two hunters hunt a stag together, but
one or both may choose instead to hunt a passing rabbit without regard
to the other player’s situation. If this occurs, the stag escapes since both
hunters are needed to catch it. Assume that if a stag is caught it provides
three units of meat to each hunter while a rabbit is not shared and provides
one unit. Also assume, as we have in all games we have seen thus far, that
the hunters never communicate their intentions with one another.

a) Create the game matrix.

Hunter B
Hunt Stag Hunt Rabbit
Hunter A Hunt Stag (3, 3) (0, 1)
Hunt Rabbit (1, 0) (1, 1)

b) Find any pure strategy Nash equilibria.

The circle-box method reveals two pure strategy Nash equilibria: (3, 3)
which corresponds to both hunting stag and (1, 1) which corresponds to
both hunting rabbit.

c) Find a mixed strategy Nash equilibrium and its corresponding expected

payoffs to each player.
Let x be the probability with which Hunter A hunts stag. Equating Hunter
B’s expected payoffs to neutralize his choice of strategy gives

3x + 0(1 − x) = 1x + 1(1 − x)

which can be solved to yield x = 31 . So Hunter A should hunt stag with

probability 13 and rabbit with probability 23 . Hunter B will have the same
mixed strategy
  since the
 game is symmetric. The expected payoff for each
hunter is 3 13 + 0 23 = 1, which we can represent as (1, 1).

d) Suppose both hunters did communicate before the hunt. Could this result
in a better outcome? Explain.
The pure strategy equilibrium (1, 1) and mixed strategy equilibrium (1, 1)
are both inferior to the payoffs if both hunters hunted stag. Therefore, if
the hunters promised each other they would only hunt stag, they would be
better off since (3, 3) corresponds to a higher payoff to both.
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 27

3.3 Pareto Optimal Outcomes

As we have seen, some equilibria are not the most beneficial outcomes to both
players; that is, there may be another outcome that improves their payoffs or at
least one player’s payoff without the other being any worse off. We say that an
outcome is non-Pareto optimal if this is the case and Pareto optimal otherwise.
If an equilibrium is not Pareto optimal, it should not be considered an acceptable
solution to a game. Consequently, some games are not able to be solved unless
negotiation or communication is allowed.

We say that a game is solvable in the strict sense if there is only one Pareto
optimal equilibrium outcome and if there is more than one, they are equivalent.
We will illustrate with two examples.

Blue
X Y
A (5, 2) (0, 4)
Red B (1, 1) (3, 3)
C (4, 0) (1, 0)
Game 3.3.1

The circle-box method can be used to find the only equilibrium in Game 3.3.1,
at (3, 3). Since there is no other outcome that better benefits one player without
hurting the other, this outcome is Pareto optimal and the game is solvable in the
strict sense.
Blue
C D
Red A (2, 2) (4, 3)
B (3, 4) (1, 1)
Game 3.3.2

In Game 3.3.2, there are two Nash equilibria in pure strategies: (3, 4) and (4,
3) which are both Pareto optimal; however, they are not equivalent. If we find
the appropriate mixed strategies, as shown in section 3.2, we get that Red should
play A with probability 43 and B with probability 14 . The same result is true for
Blue, since the game is symmetric. This corresponds to an expected value of 2.5
for each player, so the mixed strategy Nash equilibrium is (2.5, 2.5). This is not
Pareto optimal, since both (3, 4) and (4, 3) give better payoffs to both players.
Thus, the game is not solvable in the strict sense.
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 28

Activity

Adapted from Rosenthal (2011, p.74)

1. Another important game in game theory is called the Chicken Game. In

this (dangerous) game, two people drive their cars towards one another.
The first to “chicken out” and swerve away earns a payoff of zero while the
other earns a payoff of 2. If both swerve at the same time, the payoff to
each is one. If neither swerve, the result is a terrible accident so the payoff
to each is -10. This game, though stupid in the literal sense, exemplifies
an interaction between people or groups where the side that does not back
down becomes victorious. If neither back down, it is a catastrophe. As such
the game can be applied to some situations in economics or politics.

a) Create the game matrix.

Player 2
Swerve Don’t Swerve
Player 1 Swerve (1, 1) (0, 2)
Don’t Swerve (2, 0) (-10, -10)

b) Find any pure strategy Nash equilibria.

Both (2, 0) and (0, 2) are Nash equilibria. These correspond to one player
swerving while the other doesn’t.

c) Are the equilibria Pareto optimal? Explain.

Both are Pareto optimal, since any other outcome would result in a lower
payoff to one of the players. For example, with (2, 0) the outcome (1, 1)
benefits Player 2 more but not Player 1.

d) Is the game solvable in the strict sense? Explain.

No, because there is more than one Pareto optimal outcome and they are
not equivalent.

e) Suppose that Player 2 were driving a motorcycle instead of driving a car

like Player 1. If the payoff for Player 2 not swerving when Player 1 swerves
is made higher, so that it is now 10, is the game solvable? Explain.
The game matrix now becomes
Player 2
Swerve Don’t Swerve
Player 1 Swerve (1, 1) (0, 10)
Don’t Swerve (2, 0) (-10, -10)
(2, 0) and (0, 10) are Nash equilibria. The game is still unsolvable since
these are both Pareto optimal outcomes and they are not equivalent.
CHAPTER 3. TWO-PERSON NONZERO-SUM GAMES 29

f) Find a mixed strategy Nash equilibrium and its corresponding expected

payoff in the unaltered Chicken Game. Comment on the meaning of the
result.
Let x be the probability of swerving. Equating expected payoffs to neutralize
the other player’s choice of strategy gives

1x + 0(1 − x) = 2x + −10(1 − x)
10
which, when solved, yields x = 11 . Since the game is symmetric, each
player
 
should
 
swerve with probability 10
11
. The value of the game is then
10 1 10
1 11 + 0 11 = 11 . Since the probability is high in favor of swerving, both
players would be best off to do so. The value of the game is very close 1,
the payoff in the matrix when both players swerve.

2. Which outcomes in the Stag Hunt game are Pareto optimal? Is the game
solvable in the strict sense? Explain.
Recall that the matrix for the Stag Hunt is

Hunter B
Hunt Stag Hunt Rabbit
Hunter A Hunt Stag (3, 3) (0, 1)
Hunt Rabbit (1, 0) (1, 1)

Although both (3, 3) and (1, 1) are Nash equilibria, only the outcome (3,3)
is Pareto optimal. The game is solvable in the strict sense since there is only
one Pareto optimal equilibrium outcome.
Glossary

circle-box method: a method of finding an equilibrium outcome in a game; for

zero-sum games, the largest column entries are circled and the smallest row entries
are boxed; a saddle point equilibrium exists if an entry is both circled and boxed;
for nonzero-sum games, the largest first number of each of the column entries is
circled and the smallest second number of each of the row entries is boxed; a Nash
equilibrium exists if any entry has the first number circled and the second number
boxed

Dominance Principle: a player should never play a dominated strategy, since

his payoffs will always be better if he played any other strategy

dominant strategy: the strategy that a player will always choose over any other
because it gives him the best possible payoffs, no matter what the other players do

dominated strategy: a strategy that will always give a player a smaller payoff
than another strategy, no matter what the other players do; a dominated strategy
should never be chosen

equilibrium outcome: an outcome from which the players would not want to
deviate, as they would only reduce their payoff; known as a saddle point in a
zero-sum game

expected payoff: the average payoff that a player can expect to receive by play-
ing a certain strategy or set of strategies with particular probabilities; calculated
by multiplying each payoff by its associated probability and adding them all up

fair game: a game that has a value of zero, so that it is not biased toward any
player

maximin: the largest of all the minimum row entries in a game matrix which
maximizes the minimum gain

maximin-minimax method: a method of finding a saddle point in a zero-sum

game where the maximin and minimax are first found; if they are equal, the out-

30
GLOSSARY 31

come is a saddle point and is therefore the solution to the game

minimax: the smallest of all the maximum row entries in a game matrix which
minimizes the maximum gain

Minimax Theorem: an important theorem in game theory which states that

every two-person zero-sum game has a solution; the solution is the value of the
game and is unique

mixed strategy: a strategy which involves playing several strategies according

to certain probabilities; used when there is no optimal pure strategy

mixed strategy Nash equilibrium: a mixed strategy that is a solution to a

nonzero-sum game

Nash equilibrium: any equilibrium outcome in a nonzero-sum game

outcome: the result when each player chooses a particular strategy

Pareto optimal outcome: an outcome for which there is no other possible out-
come in the game that improves any player’s payoff without other players being
worse off

payoff: the value of the outcome to a player; typically, a higher number corre-
sponds to a better payoff while a lower number corresponds to a worse payoff

players: those involved in playing a game; may be individuals, companies, polit-

ical parties, etc.

pure strategy: a single strategy that a player will always choose because it gives
the best possible outcome no matter what the other players do

pure strategy Nash equilibrium: a pure strategy that is a possible solution

to a nonzero-sum game; corresponds to a saddle point in a zero-sum game

rational player: a player who keeps his own best interests in mind, ignoring how
the other player might be affected by his decision

saddle point: in a zero-sum game, an outcome that is simultaneously a row

minimum and a column maximum; a game may not have a saddle point, but if it
does, it is the solution to the game

solution: in game theory, a solution is the set of strategies that should be cho-
sen by each player to give them the best possible payoff given the situation; the
GLOSSARY 32

solution should also state the value of the game

solvable in the strict sense: refers to a game that has only one Pareto optimal
equilibrium outcome and if there is more than one, they are equivalent

subgame: a part of a game, so that one or more strategies have been removed
from the original game

symmetric game: a game for which if the labels for the row player and column
player were switched, each individual’s payoffs would not be affected

value of the game: the number v which corresponds to the row player’s average
guaranteed payoff by playing a certain strategy so that the column player’s choice
of strategy gives an average guaranteed payoff of −v.
Annotated Bibliography

1. Game Theory. (2012, May 2). In Wikipedia. http://en.wikipedia.org/

wiki/Game_theory
This source provides very detailed information on the background and his-
tory of game theory as well as some common types of games.

2. Spaniel, W. (2012). Game Theory 101. Retrieved from

http://www.gametheory101.com
This website contains a number of videos that explain the basics of game
theory. The videos would support many of the lessons in this course well.

3. Rosenthal, Edward C. The Complete Idiot’s Guide to Game Theory, Alpha:

New York, 2011. This source provides information on the importance of
game theory in our society. Although not covered in the lessons, it discusses
how factors such as threats, cooperation among players and bargaining affect
decisions in game theory. The book uses language that is easily understood
by high school students.

4. Straffin, P. D., Game Theory and Strategy, The Mathematical Association

of America: Washington, 1993.
This source provides a good introduction to game theory, including defini-
tions and examples. It also provides information on how to solve games with
more than two people.

5. Williams, J. D. The Compleat Strategyst, Dover Publications: Mineola, 1986.

This source contains more examples and stories of game theory applied
in real situations. There are an abundance of exercises good for general
practice, although they are simply game matrices with numbers and no
underlying story or situation.

33